Advancements in modern AIs, especially large language models (LLMs), have stirred discussions about the potential of AI bots to emulate, assist, or even outperform humans in various tasks such as writing essays, taking the SAT, writing computer programs, solving math problems, or developing ideas. Debates arise about their potential impact on labor markets and broader societal implications. Many of these present and future roles for AI involve decision-making and strategic interactions with humans. It is therefore critical to understand AI’s behavioral tendencies before it can be trusted with pilot or co-pilot positions in societal contexts, especially since the models and the training processes are complex and not transparent. Current evaluation and alignment of LLMs are predominantly focused on their textual outputs rather than their behaviors, which highlights a significant gap in understanding the full capabilities and limitations of these AI systems.
In response to these evolving challenges and opportunities, this article collection calls for broad discussions on the concept of AI behavioral science, which represents an emerging field that seeks to understand, model, and direct how AI behaves. Critical questions include: Do AIs have personalities? How to describe the patterns of AI behaviors? How to quantify the similarity between AI and humans behaviorally? How to conceal the objectives of AI (rather than generating the next words) and align them with the distribution of human objectives? How to model and optimize human-AI collaboration? What are the unique challenges in AI behavioral studies (e.g., sensitivity in prompting)? What is the key difference between AI behavioral science and human behavioral science? Do we need to design new experiment methodologies and measurements tailored for AI? What could be the potential applications (e.g., AI agents)?
This Research Topic aims to create a collaborative and interdisciplinary platform that brings together researchers from different fields, especially generative AI, data mining, and behavioral sciences to discuss these questions. By fostering an open and forward-looking environment, our ultimate goal is to facilitate discussions on the current landscape of AI behavioral science at large. This article collection provides an opportunity for participants to share insights, exchange ideas, and explore innovative approaches in the field.
We encourage paper submissions relevant to (but not limited to) the following topics:
● Insights from comparing AI behavior with human behavior.
● Game theoretical approaches to human-AI interaction.
● Empirical or theoretical analysis of the objectives, utilities, and rationality of AI.
● New developments in aligning LLMs with diverse human behaviors, preferences, and objectives.
● New experimental designs and measurements for AI behavioral science research.
● Applications of AI to investigating human behaviors, including but not limited to economics, psychology, sociology, education, healthcare, and the future of work.
● Ethics of AI behavioral science, including but not limited to studies that promote fairness, diversity, and representation, benefiting individuals, groups, and society at large.
Topic Editor Qiaozhu Mei is a visiting faculty researcher with Google Deepmind and is serving as a scientific advisor for MobLab Inc. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Keywords: Artificial Intelligence, Large Language Models, Behavioral Science, AI Behavior, Human Behavior
Important note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
